Time-Frequency Distribution Map-Based Convolutional Neural Network (CNN) Model for Underwater Pipeline Leakage Detection Using Acoustic Signals

• Main Authors: Yingchun Xie, Yucheng Xiao, Xuyan Liu, Guijie Liu, Weixiong Jiang, Jin Qin
• Format: Article
• Language: English
• Published: MDPI AG 2020-09-01
• Series: Sensors
• Subjects: acoustic leak signal; hydrophone; fault diagnosis; time-frequency image; EEMD; CNN
• Online Access: https://www.mdpi.com/1424-8220/20/18/5040

Detection technology of underwater pipeline leakage plays an important role in the subsea production system. In this paper, a new method based on the acoustic leak signal collected by a hydrophone is proposed to detect pipeline leakage in the subsea production system. Through the pipeline leakage test, it is found that the radiation noise is a continuous spectrum of the medium and high-frequency noise. Both the increase in pipe pressure and the diameter of the leak hole will narrow the spectral structure and shift the spectrum center towards the low frequencies. Under the same condition, the pipe pressure has a greater impact on the noise; every 0.05 MPa increase in the pressure, the radiation sound pressure level increases by 6-7 dB. The time-frequency images were obtained by processing the acoustic signals using the Ensemble Empirical Mode Decomposition (EEMD) and Hilbert–Huang transform (HHT), and fed into a two-layer Convolutional Neural Network (CNN) for leakage detection. The results show that CNN can correctly identify the degree of pipeline leakage. Hence, the proposed method provides a new approach for the detection of pipeline leakage in underwater engineering applications.